Abstract

Performance evaluation facilitates tracking and controlling project progress. Project control consists of two main steps: measurement and decision-making. In the measurement step, key performance indicators (KPIs) are designed to evaluate a project’s different aspects and are used as a thermometer to determine the health status of the project. In the decision-making step project performance is forecasted and analyzed to support needed management actions. While considerable work is available on the quantitative performance of projects, less attention is directed to qualitative performance. This research presents a framework for qualitative measurement, prediction, and optimization of construction project performance to enhance the progress reporting process and to support management in taking corrective actions, if needed. The framework has three newly developed models; KPI prediction model, performance indicator (PI) prediction model and performance optimization model (POM). The framework is developed for performance measurement, prediction, and optimization of construction projects based on six selected KPIs (cost, time, quality, safety, client satisfaction, and project team satisfaction). The selection is based on the results of a questionnaire and the literature review. Qualitative data of KPIs was collected from 119 construction projects and were then utilized in the development of the three models.
The first model maps the KPIs of three critical project stages to the whole project KPIs, based on soft computing methods. Three different soft computing techniques are studied for this purpose and their results are compared: the neuro-fuzzy technique, using Fuzzy C-means algorithm (FCM), and subtractive clustering, and artificial neural networks (ANN). The neuro-fuzzy model is developed for predicting the KPIs of the next stages of a project. The second model used the forecasted results of the first model to generate a single composite PI expressing the health status of the project. The relative weight of each KPI used in calculating the project PI is determined using the Analytic Hierarchy Process (AHP) and Genetic Algorithm (GA).
Performance Optimization Model (POM) is the third model. It is used for selecting suitable corrective actions considering the project status expressed by the six KPIs stated above. The developed model can be applied in the initial and middle stage of the project to assist owners in the improvement of the overall project PI and in the improvement of individual KPIs. Different possible modes are considered for project activities based on different ways, referee to here as modes, for resource allocation, execution methods, and/or choice of different materials. GA is applied to choose among different activity modes and optimize project performance using POM. The number of activities and their modes are flexible and do not have any limitations. MATLAB software is used for developing the models in this research. The developed framework and its three models are expected to assist owners and their agents in managing their project effectively.
Validation was conducted by using the data from 16 real projects to confirm the model’s effectiveness and to compare the results of the soft computing techniques. These results indicate that a neuro-fuzzy technique using subtractive clustering performs better than both the neuro-fuzzy technique with FCM and ANN in predicting project KPIs. The automated framework employs a set of performance indicators to evaluate, predict, and optimize the construction project’s performance, qualitatively. It applies different soft computing techniques and compares their results to choose the best technique. The developed framework can be used in construction projects to help decision-makers evaluate and improve the performance of their projects.